Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Computers that use light instead of circuits to perform calculations may sound like a plot point from an episode of Star Trek, but researchers have been working on this new approach to computing for years.
They’re called optical computers, and labs around the world are exploring how they could be useful in everyday life.
A team of researchers from Penn State published on Wednesday paper In the journal Science Advances studies how optical computing can reduce the power consumption of artificial intelligence systems.
Shengji niThe Penn State engineering professor and one of the paper’s authors told CNET that this work is a proof of concept for how optical computing could benefit the fast-growing artificial intelligence industry in the future.
“Sometimes progress comes from rethinking familiar physics with a new goal,” Ni said. “By revisiting classic ideas in optics through the lens of modern AI challenges, we can open up new practical directions for faster, greener computing devices.”
Run artificial intelligence
With the increasing adoption of AI in business and home use, the issue of AI energy costs has become relevant. A lot of computing power is needed to run AI products and services like ChatGPT, and a lot of energy is consumed in the process.
You may live in or near a town that a technology company is planning to build Data centerOr your monthly utility bill may rise due to increased demand on the local power grid.
International Energy Agency Estimates Data centers accounted for about 1.5% of global energy consumption in 2024 and that this number rose by 12% annually in the previous five years. Also the International Energy Agency Estimates Energy use in data centers could double by 2030.
For this reason, using an alternative computational method to reduce the energy consumed by AI is an attractive prospect.
Speed of light
Optical computers — computers that use light instead of electricity — are still mostly in the technology industry category, where they are still years out of commercial use. They were understood Since the sixtiesThe roots of optical information processing extend much further.
Real optical computers have mostly been moved to research laboratories. But optical data transfer, which transmits data quickly via pulses of light, is used today in some large data centers and in ground-to-aircraft transfers.
However, the use of optical computing in artificial intelligence is an emerging field of study. There are real challenges in getting light to cooperate so it can perform the functions required by neural networks, a subset of artificial intelligence used in products like chatbots today.
Basically, light naturally moves in a straight line. To build a computer that can process data, you need an optical system that produces nonlinear functions. In order for optical computers to do this, they often require other materials that can be difficult to manufacture and consume a lot of energy.
“True optical nonlinearities are usually weak and difficult to achieve – often requiring high-powered lasers or specialized materials, which increases complexity and can undermine the energy-efficiency advantage of optics,” Ni said. “Our approach avoids these requirements while still providing performance similar to nonlinear digital networks.”
Infinity mirror
Researchers at Penn State have found an interesting solution that could help optical computers perform nonlinear functions better suited to the type of data processing needed by artificial intelligence.
The prototype the team designed uses an “infinity mirror” setup that loops “tiny optical elements, and encodes data directly into the rays of light,” creating a non-linear relationship over time. The light patterns are then captured with a microscopic camera.
“The main takeaway is that a carefully designed optical structure can produce the nonlinear input-output behavior that AI needs without relying on strong nonlinear materials or high-power lasers,” Ni said. “By allowing the light to resonate through the system, we create these nonlinear maps while keeping the hardware simple, low power, and fast.”
The figure (above) shows how light is focused into a small processing unit, allowing vast sequences of computational information to be transmitted without using power-intensive circuits. The other figure (below) shows how the team process works in theory. The input of light is repeatedly reflected by lenses and other optical devices, encoded by complex strings of information, and finally focused into a camera that provides a streamlined output.
It’s an interesting concept, but turning a prototype into a system with real-world applications will take a lot of time, work, and money.
From laboratory to data center
Ni admits that we are still a long way from optical computers with artificial intelligence.
“A realistic timeline for an industry-facing prototype and early demonstrations is between two and five years, depending on the level of investment and target application,” he said.
However, it is a hot topic in the computing world. Francesca ParmigianiThe principal research director at Microsoft Research told CNET that optical chips could one day work alongside traditional graphics processing units to help AI systems perform specific tasks.
“Optical computing has the ability to perform a much larger number of operations in parallel and at much higher speeds than traditional digital devices,” Parmigiani said. “This can translate into significant gains in energy efficiency and reduced latency for workloads.”
The traditional computers we use for AI will not be replaced by visual computers any time soon. But within a few years, it is possible to integrate optical computers into artificial intelligence systems to work with regular computers.
“The goal is a hybrid approach: electronics still handles general-purpose computing, memory and control, while optics can accelerate specific, large-scale operations that dominate the time and energy cost of AI,” Ni said.